Latest Articles for Training Data

A study on improving training efficiency for language models using SlimPajama dataset.

A new approach enhances the training of surrogate models in software development.

This study examines how tokenization impacts gender bias in translation models.

This method helps neural networks avoid local minima and learn more effectively.

Two advanced language models for Modern and Rabbinic Hebrew are now available.

Examining how machine learning models improve predictions in materials research.

Introducing MetaCLIP for better image-text data collection.

New methods improve image editing speed and quality using smaller models.

A new model helps enhance recommendations by addressing noisy user feedback.

A new approach enhances commit message quality by leveraging best practices.

A look at how machine translation reflects and reinforces gender bias.

CleanSheet advances model hijacking without altering training processes.

Examining how deep neural networks learn and the challenges they face.

New models aim to improve language tech for Brazilian Portuguese speakers.

New unbounded language model improves predictions using extensive data.

Examining harm amplification in text-to-image models and its societal impact.

WiOpen efficiently recognizes both known and unknown gestures using Wi-Fi technology.

A new open language model for research and innovation in natural language processing.

Examining difficulties in recognizing languages in mixed-language communication.

Learn how animated stickers are made from text and images.

A new approach helps AI measure uncertainty and improve decision-making accuracy.

Understanding unlearnable example attacks through game theory for better data protection.

A study on enhancing language model learning using minimal style changes in training data.

This article examines how language models can adopt ideological biases from training data.

Examining biases and rationality in large language models used for financial analysis.

A critical look at the true capabilities of Generative Adversarial Networks.

Examining grokking in deep learning and its implications for performance.

Study explores Elo rating system's impact on medical student learning.

MobiLlama offers efficient language processing for devices with limited resources.

RoadRunner helps robots navigate difficult outdoor terrains safely and efficiently.

New datasets improve depth estimation models for various environments.

Examining how to fairly reward artists in the age of AI-generated art.

Research shows how LLMs can expose training data, raising privacy concerns.

The Yi model family showcases strong language and multimodal processing capabilities.

New training framework enhances language model learning through structured data.

A novel approach to find backdoor samples without needing clean data.

New tools are helping scientists predict protein stability and its implications for health.

A two-stage method enhances model performance across different data groups.

New methods improve accuracy in predicting protein-ligand interactions.

A guide to understanding music similarity in generative models.